---
title: Agentic Business Process Management (APM)
type: concept
tags: [apm, bpm, agentic-ai, paradigm]
sources: ["[[sources/2026-calvanese-agentic-bpm-manifesto]]", "[[sources/2025-calvanese-autonomy-business-process-execution]]", "[[sources/2025-elyasaf-self-modifying-abps]]", "[[sources/2025-fournier-agentic-ai-process-observability]]", "[[sources/2026-theodorakopoulos-bi-bpm-genai-review]]"]
created: 2026-04-13
updated: 2026-05-11
---

# Agentic Business Process Management (APM)

**Definition 2.1** ([[sources/2026-calvanese-agentic-bpm-manifesto]]): *An Agentic Business Process Management (APM) system is an agentic socio-technical system, jointly realized by a collection of agents, some of which are at least partially **process-aware** agents.*

APM extends [[concepts/process-awareness|BPM]] to treat **software and human agents as primary functional entities** that perceive, reason, and act within explicit process frames. It marks a paradigm shift from **automation-oriented** BPM/RPA systems (which execute predefined workflows) toward **autonomy-oriented** systems in which goal-directed agents operate under [[concepts/framed-autonomy|framed autonomy]].

## Key distinctions

- **Autonomy ≠ automation.** Automation = predefined execution; autonomy = perceive/reason/choose under a frame.
- **APM ≠ AI-Augmented BPMS.** [[sources/2023-dumas-ai-augmented-bpms|ABPMS (Dumas et al. 2023)]] embeds AI into BPM functionalities but leaves agent-centricity implicit. APM makes agents first-class citizens.
- **APM ≠ MAS.** Classical multi-agent systems research focuses on allocating agents to tasks within a known process; APM focuses on managing agents as first-class citizens ensuring achievement of business goals.
- **APM extends ABPMS.** Where [[sources/2023-dumas-ai-augmented-bpms|ABPMS (2023)]] embeds AI functionalities into a BPMS, APM additionally treats agents as first-class citizens with framed autonomy and process-awareness — see [[syntheses/abpms-to-apm-evolution]].

## Agent categories in an APM
- **Human agents** — process workers, managers.
- **Software agents** — code-executed, without deliberation.
- **Embodied agents** — robots acting on the physical world.
- **AI agents** — deliberation via AI models (e.g., LLMs).

## Architecture (two levels)
- **Macro / APM System:** the "world" where agents interact via tools and protocols (MCP, ACP, RDF, KQML); framing is imposed here.
- **Micro / APM Agent:** individual [[concepts/perceive-reason-act|Perceive–Reason–Act]] loop with perception, reasoning, and action modules.

## Required agent capabilities (ordered)
1. [[concepts/framed-autonomy|Framed autonomy]]
2. [[concepts/explainability-apm|Explainability]]
3. [[concepts/conversational-actionability|Conversational actionability]]
4. [[concepts/self-modification|Self-modification]]

## PMAI'25 / AutoBiz Dagstuhl follow-ups
Three 2025 workshop papers refine the manifesto capability stack:
- [[sources/2025-calvanese-autonomy-business-process-execution]] — elevates goals and [[concepts/normative-frame|normative frames]] to first-class abstractions.
- [[sources/2025-elyasaf-self-modifying-abps]] — structures the [[concepts/self-modification]] problem space; proposes [[concepts/abps-autonomy-levels]].
- [[sources/2025-fournier-agentic-ai-process-observability]] — operationalises observability over LLM-agent trajectories via [[concepts/agent-process-observability]].

## Adjacent paradigm: Large Process Models (Kampik et al. 2024)
[[sources/2024-kampik-large-process-models|Kampik et al.'s LPM vision]] is a parallel paradigm centred on a process-fine-tuned LLM in a [[concepts/neuro-symbolic-bpm|neuro-symbolic stack]] (knowledge graphs, process atoms, classical BPM tooling). APM and LPM agree on neuro-symbolic hybridisation and disagree on the centre of gravity: LPM centres on the LLM with symbolic scaffolding; APM centres on goal-directed agents with LLMs as one reasoning substrate. See [[syntheses/llm-bpm-reading-list]] for the joint reading list. Concrete instantiations:
- [[sources/2026-licardo-bpmn-assistant]] — BPMN Assistant: LLM + JSON IR + atomic editing functions for BPMN modelling.
- [[sources/2025-varsani-neuro-symbolic-ai-sap-erp]] — SAP ERP grounding: LLM + middleware + symbolic rule engine for natural-language ERP queries.

## Adjacent framing: 5-layer integrative architecture (Theodorakopoulos & Theodoropoulou 2026)

[[sources/2026-theodorakopoulos-bi-bpm-genai-review|Theodorakopoulos & Theodoropoulou 2026]] propose a parallel **5-layer integrative framework** — Data & Computation → Organisational Insight → Process Intelligence → Augmentation → Decision — for connecting BI, BPM, big data analytics, GenAI, and DSS. GenAI is positioned explicitly as the **Augmentation layer**, mediating between layers but *not* the new core: derivative in the sense that "augmentation is only as strong as the evidentiary, process, and governance foundations it augments."

Compared to the APM Manifesto's agent-centric ordered-capability stack (framed autonomy → explainability → conversational actionability → self-modification), the Theodorakopoulos framework is **technology-stack-centric** rather than **agent-centric**. The two are complementary rather than competing — APM tells us *who acts* under what frame; Theodorakopoulos tells us *what layers of capability* an APM system rests on. The latter is more cautious about GenAI's transformative potential.

## Redesign-heuristic implications (2026)
The agentic-extension of the [[concepts/bpr-heuristics|Reijers & Liman Mansar (2005) catalogue]] called for by [[sources/2026-dumas-agentic-bpms-pyramid|Dumas et al. (2026, p. 45)]] is filed at [[syntheses/ai-agent-redesign-heuristics-2026]] with a companion pedagogical HTML map ([[assets/ai-agent-heuristics-devils-quadrangle.html]]). The synthesis re-reads the 29 classical heuristics and adds 10 brand-new ones rooted in the APM capability stack (framed autonomy → explainability → conversational actionability → self-modification).

## Related concepts
[[concepts/framed-autonomy]] · [[concepts/normative-frame]] · [[concepts/process-awareness]] · [[concepts/perceive-reason-act]] · [[concepts/self-modification]] · [[concepts/abps-autonomy-levels]] · [[concepts/agent-process-observability]] · [[concepts/llm-assisted-process-modelling]] · [[concepts/neuro-symbolic-bpm]] · [[concepts/bpr-heuristics]] · [[concepts/devils-quadrangle]]